Excuse me, my study is talking about the corporate governance and firm performance with use some of moderating variables. I use the ROA as a measurement for performance. But some of the firms have a Negative ROA. I use panel technique.
Treat the negative data in the set as part of the values of the set. There is no need to give it special treatment. However, it/they should not be eliminated if it/they had been included in the original sample.
If you expect the value to be positive and it turns out to be negative (which may contravene your performance-governance hypothesis)---you need to verify their frequency, distribution, etc. whether they are frequent in number so as to present statistically significant challenge to the current literature.
If your hypothesis is: "ROA indicates firm's performance for governance study" and if firm A meets all criteria for "good governance" but shows negative ROA---this raises a challenge to the proposed hypothesis. Thus, look further into the source of causality of the negative ROA and its relevance to governance.
Some new discovery generally come from what we do not expect. Pursue it further; don't ignore or remove it. Look at it as a potential opportunity for new discovery. Cheers.
http://www.nber.org/papers/w14851.pdf
Hassan I think you can convert that series in to positive numbers. For this you need to identify smallest number in the series and add the same number in whole series. As you are checking the impact your result wont be affected.
Dear Mohammad,
I am not sure whether you should use absolute numbers by converting the negative observations to positive values. Performance can be negative, thus the reasons have to be investigated. We should not change our observations in order to "make" them match our hypotheses.
Paul
- Badges
- Science topic
- Similar topics
- Biological Science
- Bioecology
- Systems Ecology
- Landscape Ecology
- Connectivity